 proud to have with me Brad McCready from IBM. He is the guy who started off Open Power and started off all of the work on the Power 8 and Power 9. So thanks very much for being here. Great to be here Dave. Good day. Great. Well can you tell me a little bit more about the announcements you've made? I read that Google had an announcement and maybe that you needed to catch up a bit. Well if Google was going to talk about Power 9 we better talk about it first or at least tell the industry that exists. So we did a series of keynotes here today where I talked about our Power Roadmap. Provided our Power 8, Power 8 with NVLink and then Power 9. Some details on that and then Google of course then did follow that up with their discussion on how they're partnering with Rackspace to build Power 9 systems on the open compute platform. And we also emphasize a lot the accelerated computing. We think that's really important here as we're entering this era of you know the demise or whatever we want to say of Moore's Law. You know in Rackspace talked about that. You know we showed we had a lot of new accelerator buses on Power 9 and Rackspace is leveraging those and some of their future development efforts as well. Well you know congratulations on what has happened to Open Power compared with 214, 215 and now it's what is it 60 announcements this today. You got them all sitting out over here you know our 59 new products. Oh 59 I'm sorry. I should have let you round up. I should have let you round up. You're right. So what all those years ago what was it that made you start this whole project of the Open Power? Oh it's we kind of talked about it you know there's there's a couple of major forces playing on the industry right now and one of the major forces is of course as we talked about the waning of Moore's Law. So Moore's Law fueled so much of the cost performance takedown that our clients need. But by Moore's Law you mean the speed part of it there's still the transistor part of it seems to be going strong as well. Well really what it is is it's an economic statement cost performance and that's what's waning you know and we had a lot of charts that were shown today and a lot of charts have been shown over the past years that you know that cost performance takedowns not there. So what are we doing? We're doing all sorts of new innovation to make up for that slack. Nobody showed up and said you know what I don't need my computer to get any lower on cost performance. They're still counting on it to be there so that's one force that led to the Open Power. And then the other force is is the consumption models of IT is changing you know whether you look at hardware, software, just about anything people are consuming you know open technologies that's the way they want to consume technology. So you know with those two forces needing lots more innovation the Open Power our 60 innovations that we've shown here is very important to have the whole industry innovating and we need to make things open so that the whole broad market segments could consume it. So earlier on Calista said that your motto is now 20 in 2020 20 percent of the marketplace in 2020. So what more do you need to do to that's a very ambitious market target what more do you think that you need to be doing in the development etc to succeed in that? Yeah so you know we need to we need to do a couple of things one is is we need to continue to provide differentiated innovation. So you know we need to give the people the tools they need the IP they need the platforms they need to innovate new things so accelerate new workloads you know what new ways to attach accelerators. Then the other thing we of course need to do is is we need to make make lots and lots of developers have access to the platform. So after we provide those innovations we got to get broad access to it. So we're doing things with both university programs as well as you know development co-development research labs to enable people to get access to these platforms. So what are some other people and myself have talked about is is bringing the silicon closer to the software so that you are trying to reduce the patholanks reduce the time to make up for Moore's law and Amdahl and escape from Amdahl's law. So how do you see that happening how what is it what's the part of open system that makes that happen? You know you know bringing the silicon closer to the platform you know maybe don't think of it is is changing the software stack or changing a different stack maybe think of it as customization you know hey I need a I need a new memory technology that has this kind of latency this kind of bandwidth this oh I want to go design something special to provide that okay let's go let you design that. I need this type of a new accelerator that has one of these one of those let's make it very quick and easy to design that. So you can bring the silicon closer you know maybe isn't a technical thing in the form of we got to change the software stack or how things work what you need to do is let people customize you know we're seeing you know this this particular conference we're here you know embedded in the GTC conference we're seeing how you know custom you know accelerators GPUs on the very important workload deep learning really a great match and you know that's bringing the silicon closer you know to the to the workload because the particular structure the GPU is very very well matched to what deep learning needs in the way of matrix multiplies etc. So one of the highlights of the power 8 was the the CAPI interface which a lot of people have worked with. NV link that's a new thing that you're bringing out can you talk more about that why is that was that necessary what does it do? Brings a silicon closer to the workload just like you said you know NV link right you know so many you know important workloads are being accelerated by GPUs but you know what the biggest problem we had was we couldn't get the data into and out of the GPU fast enough so what we need to do well we need to build a special piece of hardware that allows the GPU to talk to the processors memory faster so we can get the you know the data in and out faster what we did we brought the you know brought the computation much much closer to the workload and that type of customization is going to have a big impact on the industry because it's a highly differentiated way notice so many people that say oh if I can move things in and out of a GPU at 80 gigabytes per second instead of five you know I'm going to get some new workloads done and that's what we hope to have happen excellent so what about the the open part of it as you said you're developing some hardware developing new pieces of hardware itself what about the software side is that an area that you're seeing collaboration and getting things done quicker or new techniques oh yeah you know we're providing that open software innovation all the way up the stack you know one of the places right now where the industry is you know really struggled with open software is down there even in the firmware level you know so many of these systems have to go into these large internet data centers and need to have special ways to boot and special ways to update the firmware and all that you know and that firmware has been closed a lot one of the big things that we've done with our barrel i processor we built a rack space as we open up all those firmware layers and you know that open software now is all available out on github and people can innovate with that you know we're all we're allowing that innovation to take place at every level and of course we have our Linux and our open stack and our kvm all of those out there and available then to get up there to the middleware you know we're now letting you know people take you know a lot of the open source databases mongo etc you know provide acceleration of that through advanced flash techniques and then you'll capi attach flash is one of the things that we're using to speed up those databases great so um if we if we come back here in two years time three years time uh 2019 what do you hope to uh what would you think is going to be different what are you hoping to see uh in that sort of timescale you know so so right now you know what we're doing you know and i did when i'm pointing to it i know all of our viewers can't see it but we have so many pieces of hardware out here um that are uh um you know being built and developed you know and that's what we're about right now we need to get that breadth of hardware development you know in two three years time scales you know i'm hoping now that by then now we'll get into that nether phase which is after you get all this base hardware built we're going to see lots and lots and lots of accelerated solutions we have a few of them right now just started but we need those accelerated solutions because that's ultimate is going to bring the cost performance down which is again our stated goal so you know we'll be shifting you know more and more from hardware and after we get that base out there and more and more into accelerated solutions on top of this hardware right so volume is key to uh to success the base it's the base right well thanks very much indeed to you and thanks to the open power foundation and congratulations on a on a fantastic show that you put on today so all the best and uh thanks very much thanks for watching